Transmembrane I~-Barrel Proteins

TRANSMEMBRANE I~-BARREL PROTEINS

By GEORG E. SCHULZ

Institute for Organic Chemistry and Biochemistry, Albert-Ludwigs-Universit~t, D-79104 Freiburg im Breisgau, Germany

I. Introduction ...... 47 II. Structures ...... 49 III. Construction Principles ...... 55 IV. Functions ...... 59 V. Folding and Stability ...... 61 VI. Channel Engineering ...... 63 VII. Conclusions ...... 65 References ...... 66

I. INTRODUCTION After numerous complete genomic DNA sequences had been established and after these had been interpreted to a great extent in terms of protein sequences, it became evident that about 20% of all proteins are located in membranes (Liu and Rost, 2001). This number was derived from a search for transmembrane 0~-helices with a computerized prediction system, the results of which are known to come with a high confidence level. Such helices can be recognized from a continuous stretch of 20 to 30 nonpolar residues with a predominance of aliphatic side chains at the center and aromatic residues at both ends (Sipos and von Heijne, 1993). The number of transmembrane a-helices per protein is broadly distributed and averages around six. The main chain amides of these a-helices are all locally complemented, so that the interface to the nonpolar membrane interior is exclusively formed by the nonpolar side chains. This explains the use- fulness of an a-helix as a membrane-crossing element. The helix orientation can be deduced from the charge patterns of the interhelical segments, positive charges being inside and negative charges outside the cell. As a consequence the coarse structures, usually called topologies, of all these a-helical membrane proteins can be assessed from the sequence. The presence of additional r-sheets in these proteins is discussed. However, these must be internal because it cannot be expected that the membrane is faced by a mixture of a-helices and r-sheets as the main chain hydrogen bond donors and acceptors at the sheet edges cannot be complemented by those of a-helices.

47 CopDight 20113, Elsevier Science ! I ;SA ADVANCES 1N All rights I ~v'l ~cd PROTEIN CHEMISTRY, VoL 63 006%3233 '0~~, $35 011 48 GEORG E. SCHULZ

Because of their construction, transmembrane/3-sheets have an amide saturation problem at their edge strands. Fusing both edges to form a barrel, however, permits complementation of all amide hydrogen bond donors and acceptors. Like a transmembrane helix, a fl-barrel can face the nonpolar interior of the membrane if its outer surface is coated with nonpolar side chains. Such barrels occur, indeed, in the outer membrane of gram-negative bacteria. They should be detectable in the sequence because every second residue is nonpolar. The significance of this information is low, however, because the individual /3-strands are only slightly more than half a dozen residues long and the inter- mittent residues pointing to the barrel interior can be both polar and nonpolar. From the start the fl-barrels require cooperative folding of a polypeptide chain of 100 or more residues which constitute an entropic hurdle. In contrast, an a-helical transmembrane protein can traverse the membrane as soon as a local segment of about 20 residues becomes nonpolar. The remaining transmembrane part can then be added piece by piece, which is entropically much more favorable. Therefore the/3-barrel membrane proteins arose probably rather late during protein structure evolu- tion, constituting an addition to the much simpler a-helical membrane proteins. At present, transmembrane/3-barrel proteins have been found exclusively in the outer membrane of gram-negative prokaryotes; these membranes seem to lack a-helical proteins. Accordingly, a separation exists between a-proteins in all cytoplasmic membranes and/%proteins in the specialized outer membranes. Following the endosymbiotic hypothesis, /3-proteins are also expected in the outer membranes of mitochondria and chloroplasts, but none of these proteins has yet been structurally elucidated. Given the limited abundance of such membranes, the/~-proteins are likely to make up only a small, special class of membrane proteins. In contrast to their limited importance in nature, the fl-barrel proteins are most prominent in the list of established membrane protein structures. Moreover, they show a high degree of internal chain-fold symmetry and therefore convey the impression of beautiful proteins (Fig. 1). One should not forget that the first protein structure, myo- globin, caused some disappointment among those who solved it as it showed no symmetry whatsoever; even the a-helices were not whole- numbered but about 3.6 residues per turn. Accordingly, the symmetric transmembrane/3-barrels stand out from the bulk of asymmetric chain folds of water-soluble proteins. The number of distinct chain folds of integral membrane proteins is probably much smaller than the respective number for water-soluble TRANSMEMBRANE fl-BARREL PROTEINS 49

3_ FIG. 1. Ribbon plot of the 16-stranded /3-barrel of the general porin from Rhodobactev"capsulatusviewed from the molecular threefold axis (Weiss and Schulz, 1992). Note the large variation of the inclination angle c~ and the difference between ~he high barrel wall facing the membrane and the low wall at the trimer interface. proteins, which ranges around a thousand (Schulz, 1981; Brenner and Levitt, 2000). Proteins of the cytoplasmic membrane consist mostly of transmembrane a-helices, and the bacterial outer membrane proteins contain fl-barrels. Both types show a high neighborhood correlation which limits the number of different topologies appreciably (Schulz and Schirmer, 1979). The a-helices run, in general, perpendicular to the membrane plane and connections are formed between neighboring helix ends (Bowie, 1999). All transmembrane fi-barrels contain mean- dering all-next-neighbor antiparallel sheets, the topologies of which are completely described by the number of strands.

II, STRUCTURES Here we discuss structures that have been established at the atomic level revealing the exact conformation of the polypeptide chain. All were determined by X-ray diffraction analysis of a three-dimensional protein crystal. Some c~-helical membrane protein structures have been analyzed by electron diffraction of two-dimensional crystals, although generally with a lower accuracy. For a long time structural analyses by NMR 50 GEORG E. SCHULZ

TABLE I t-Barrels in Water-Soluble Proteins

n S R(A) 0t(°) Secondary structure LflH proteins 1 -18 -- -- Left-handed t-helix Metalloprotease 1 18 -- Right-handed t-helix Pectate lyase 1 20 -- -- Right-handed t-helix Chymotrypsin 6 8 6.5 45 Symmetric anfiparallel t-sheet TIM 8 8 7.2 37 All-next-neighbor parallel t-sheet Streptavidin 8 10 7.9 43 All-next-neighbor anfiparallel t-sheet Lipocalins 8 12 8.7 48 All-next-neighbor antiparallel t-sheet had been considered inapplicable to membrane proteins, because these proteins can only be solubilized in large micelles. However, two small membrane proteins have now been tackled with this method, and success has been reported (Arora et al., 2001; Fernandez et al., 2001a,b). Moreover, the native structure of gramicidin A in the membrane was deduced from solid-state NMR data (Ketchem et al., 1993). Other methods providing important structural information on proteins are light spectroscopy, electron microscopy, and atomic force microscopy. However, none of them reaches atomic resolution. The existence of t-barrels was established for chymotrypsin at a very early stage in the now common protein crystal structure analyses. This enzyme contains two distorted six-stranded /3-barrels with identical topologies (Birktoft and Blow, 1972). A selection of/3-barrels in water-soluble proteins is given in Table I. The very abundant TIM-barrel consisting of eight parallel /3-strands was also detected rather early (Banner et al., 1975). Additional eight-stranded/3-barrels of this group are those of streptavidin (Hendrickson et al., 1989) and of the lipocalins (Newcomer et al., 1984). The/3-helices also belong to this group as they can be taken as single- stranded/3-barrels (n = 1) with large shear numbers of S = 18 and more (Table I, Fig. 2). They were first detected with pectate lyase (Yoder et al., 1993). The fight-handed and left-handed versions have positive and negative S values, respectively. The cross-sections of these/3-helical barrels deviate drastically from circles. The cross-section of the pectate lyase type resembles a boomerang and those of the metalloproteases a flat ellipse (Baumann et al., 1993), whereas the LflH (left-handed/3-helix) types of proteins form triangles (Raetz and Roderick, 1995). X-Ray diffraction analysis is a suitable and convenient method for obtaining exact structures of membrane proteins, but it requires three-dimensional crystals. Membrane protein crystallization has always TRANSMEMBRANE B-BARREL PROTEINS 51

A B 2rcR R 'j -# S gl~ll • I i 7-. "" ,' I 3 : ! ! I a " " ' ,,'i ? i t 1 / ,", / / [ / [ / / / b]~- [ ',

b 1 n ".o 2 1 n

FIG. 2. General architecture of a/Lbarrel. The description depends neither on the sequence of the strands nor on their directions. Residues are represented by their Ca atoms. (A) Barrel geometry with the tilt angle a of the t-strands versus the barrel axis. The hydrogen bonding pattern is shown for antiparallel t-strands. (B) The barrel is cut where the first strand reaches the upper end, flattened out, and viewed from the outside. The sequence of the n strands along the circumference is arbitrary. The thin line follows a pleat of the sheet, that is, the hydrogen bonds. The shear number S is derived by running from a given strand to the left along the hydrogen bonds once around the barrel and counting the residue number S to the point of return to the same strand (Murzin et al., 1994a,b; Liu, 1998). The depicted t-strand tilt corresponds to a positive S-value, and a tilt to the left to a negative S. been a bottleneck. Part of this obstacle is the preparation of suffi- cient homogeneous membrane protein material, because the limited volume of the two-dimensional entity membrane cannot incorporate large amounts of a recombinant protein. Moreover, any tampering with the membrane is highly hazardous for the respective organism so that high expression levels are generally rare. This problem was circumvented by expressing membrane proteins into the cytosol and (re)naturing it from there into micelles (Schmid et al., 1996), which is possible for a number of t-barrel proteins. As a general observation the crystallization of the bacterial outer membrane proteins appears to be easier than that of the u-helical proteins from the plasma membrane. Accordingly, the list of the structurally established t-barrel membrane proteins is comparatively long. It is given in Table II. The resolution of the analyses ranges from 1.6 A for OmpA (neglecting the nonnative gramicidin-A crystals) to 3.2 it for the porin OmpC (OmpK36). The crystals are usually loosely packed, except for one crystal form of OmpA that reached 50% (v/v) protein in the crystal but diffracted merely to medium resolution (Pautsch and Schulz, 2000). 52 GEORG E. SCHULZ

TABLE II Transmembrane fl-Barrel Proteins ~

n S R(A) b 01 (°) b Oligomer Reference

Gramicidin A 1 6 3.4 77 Dimer Ketchem et al., 1993 (native) Gramicidin A 2 -6 Cs+ fits -- Dimer Wallace and (CsC1, CH3OH) ~ Ravikumar, 1988 Gramicidin A 2 -4 narrow -- Dimer Langs, 1988 (C6H6, ethanol) c Nanotube 1 '8' 5.6 90 Stack Ghadiri et al., 1994 OmpX 8 8 7.2 37 Mono Vogt and Schulz, 1999 OmpA 8 10 7.9 43 Mono Pautsch and Schulz, 1998 OmpT 10 12 9.5 42 Mono Vandeputte-Rutten et al., 2001 OmpLA d 12 12 10.6 37 Mono Snijder et al., 1999 TolC e 12 '20' 13.6 51 "Trimer" Koronakis et al., 2000 a-Hemolysine 14 '14' 12.3 37 "Hepta" Song et al., 1996 Porin R. capsulatus 16 20 15.5 43 Trimer Weiss et al., 1990 Porin OmpF 16 20 15.5 43 Trimer Cowan et al., 1992 (PhoE, OmpC) Porin R_ blasticus 16 20 15.5 43 Trimer Kreusch and Schulz, 1994 Porin P denitrificans 16 20 15.5 43 Trimer Hirsch et al., 1997 Porin Omp32 16 20 15.5 43 Trimer Zeth et al., 2000 Maltoporin (two 18 20 17.1 40 Trimer Schirmer et al., 1995 species) Sucrose porin 18 20 17.1 40 Trimer Forst et al., 1998 FhuA 22 24 19.9 39 Mono Locher et al., 1998 FepA 22 24 19.9 39 Mono Buchanan et al., 1999

a All sheets are antiparallel except for native gramicidin A. The topologies are always all-next-neighbor. b The radius is calculated for a circular cross section. The angle ct can vary by 4-15 ° around the barrel. c Nonnative conformations of short peptides depend on the crystallization condition. d This enzyme exists as a monomer in the membrane and becomes active on dimerization. e The barrel itself consists of parts from several subunits, violating the rule of one barrel per subunit.

The first data showing a transmembrane barrel came from electron microscopy and electron diffraction of two-dimensional porin crystals (Jap, 1989). They followed spectroscopic studies of the amide bands in the infrared range indicating that the porin contains t-strands tilted at a 45 ° angle against the membrane plane (Nabredyk et al., 1988). As TRANSMEMBRANE/~-BARRELPROTEINS 53 can be visualized in Fig. 2A, a 45 ° tilt in a barrel can point either to the right or to the left, the two alternatives being mirror images. This ambiguity is resolved, however, because only one of them corresponds to the energetically favorable and therefore commonly observed/3-sheet twist. In conjunction with the barrel diameter known from the electron diffraction data, these data would have sufficed for modeling the/3-barrel of the analyzed porin in a rather detailed manner. Shortly thereafter an X-ray analysis of three-dimensional porin crystals clarified this structure in atomic detail (Weiss et al., 1990). The pentadecameric antibiotic peptide gramicidin A forms channels through membranes that allow the passage of alkali ions. It has been included in Table II because it forms fl-barrels in two artificial conformations as well as in its native conformation. The artificial conformations were obtained by crystallizing from nonpolar solvents in the presence and absence of Cs + ions (Wallace and Ravikumar, 1988; Langs, 1988). They showed very narrow fl-barrels, indicating that they present artifacts instead of channels. The real structure of gramicidin A consists of a somewhat wider barrel in the membrane, which could only be determined by solid-state NMR of membranes containing gramicidin A (Ketchem et al., 1993). It revealed a helical n = 1 type of/3-barrel, two of which associate head-to-head forming a channel through the membrane. Such narrow barrels can only be assumed if L-amino acid residues alternate with D-amino acids (or glycines) along the peptide chain, and this is here actually the case. The artificial nanotubes listed in Table II follow the design of the gramicidin A channel except that the chain is an eight-membered ring instead of a fl-helix (Ghadiri et al., 1994). The rings are stacked forming a channel. The smallest established transmembrane /~-barrel of the canonical type is that of OmpX (Vogt and Schulz, 1999). It contains eight strands with a shear number S = 8 and appears to represent the minimum construction for a transmembrane fl-protein (Table II). In contrast to OmpX, the ubiquitous outer membrane protein A pos- sesses an N-terminal 171-residue domain (here called OmpA) as a lnem- brane anchor and a C-terminal periplasmic domain binding to tile peptidoglycan cell wall. OmpA contains an eight-stranded ¢3-barrel with a shear number S = 10, which is larger than that of OmpX, giving rise to a larger barrel cross-section (Pautsch and Schulz, 1998). The relationship between strand numbers, shear numbers, and barrel radii is illustrated in Fig. 3. Applying point mutations, the OmpA crystals could be improved to diffract to 1.6 A resolution (Pautsch and Schulz, 2000). This allowed for an anisotropic structure refinement revealing the major mobility directions of loops and turns. It demonstrated that the turns and loops 54 GEORG E. SCHULZ

[h]l

2O S

" / 12 16 54 ° ! 8 0 4 a 4~ 32°

2O ~

Fro. 3. The observed fl-barrels (no fl-helices included) concentrate at tilt angles a between 35 ° and 50 °. Because of the two-residue repeat in the hydrogen bond pattern (Fig. ~A), the shear number S is always even. The radius R of the barrel increases with the number of strands n as well as with the shear number S. Up to OmpIA (n = S = 12) the interior of the barrel can be filled by side chains and fixed water molecules. have asymmetric mobilities which correspond to those of a model in which the polypeptide is replaced by a resilient wire. The bacterial outer membranes also contain some enzymes, among them the protease OmpT consisting of a 10-stranded/~-barrel with a shear number S = 12 (Vandeputte-Rutten et al., 2001). While the "lower" part of the r-barrel is immersed in the membrane as usual, its "upper" part protrudes into the external medium and contains the catalytic center where foreign proteins are split. OmpT is of medical interest because it contributes to the pathogenicity of bacteria. A further surfacial enzyme is the phospholipase A (OmpLA) that destroys lipopolysaccharides. It consists of a 12-stranded r-barrel with a shear number S = 12 (Snijder et al., 1999). Its barrel contains a solid interior hydrogen bonding network without a pore, a nonpolar outer surface, and the catalytic center at the external end. OmpLA is active as a dimer accommodated in the membrane. Interestingly, the dimer interface is for the most part nonpolar. The r-barrels of TolC (Koronakis et al., 2000) and of a-hemolysin (Song et al., 1996) are special because they are composed of several pieces coming from different subunits (Table II). The TolC barrel consists of three four-stranded all-next-neighbor antiparallel r-sheet pieces coming TRANSMEMBRANE B-BARREL PROTEINS ~:~ from the three subunits. The major parts of the subunits are a-helical and not in the outer membrane. The shear number S is as large as 20, giving rise to a wide channel along the barrel axis (see Fig. 3). This is in contrast with the 12-stranded t-barrel of OmpLA that has a smaller S and a solid core. The a-hemolysin barrel consists of seven t-hairpin loops coming from the seven subunits. Each subunit is water-soluble. The t-hairpin loop undergoes a large conformational change during the cooperative process of t-barrel formation, which is likely to occur on membrane insertion and after the large globular parts have formed an annular heptamer (Olson et al., 1999). The porins in the outer membrane of gram-negative bacteria are passive channels showing various grades of selectivity. The most common type consists of 16-stranded t-barrels with a shear number S = 20. The structure of the porin from Rhodobacter capsulatus illustrated in Fig. 1 was the first transmembrane t-barrel structure to be established at atomic resolution (Weiss et al., 1990). It revealed the general construction principles (Schulz, 1992, 1994), which were subsequently also observed in other membrane proteins (e.g., the aromatic girdles), in other transmembrane t-barrels (e.g., the short periplasmic turns), and in other porins of this type (e.g., the transversal electric field for polarity separation). The structures of two channels that are highly selective for maltooligosaccharides and sucrose, respectively, showed 18-stranded barrels with kidney-shaped cross sections (Schirmer et al., 1995; Meyer et al., 1997; Forst et al., 1999). These cross sections deviate strongly from circles and allow long narrow channels which are required for the selection process. The largest t-barrels have been observed with the monomeric iron transporter proteins FhuA and FepA. The structure of FhuA was established independently by two groups (Locher et al., 1998; Ferguson et al., 1998). It is known with and without a ligated siderophore. The structure of the ferric enterobactin receptor FepA is homologous to that of FhuA showing identical topology and a similar transport mechanism (Buchanan et al., 1999). In both cases there are more than 700 residues assembled in two domains: an N-terminal 150-residue domain is located inside a C-terminal 22-stranded t-barrel with a shear number S = 24.

III. CONSTRUCTIONPRINCIPLES The general structural properties of fl-barrels are illustrated in Fig. 2. The barrel is sketched as a cylinder with circular cross section and all fl-strands are assumed to run at the same inclination angle a. The fl-pleated sheet parameters a --- 3.3 A and b = 4.4 A refer to all kinds of 56 GEORGE. SCHULZ fl-sheets: parallel, antiparallel, or mixed. The hydrogen bonds in Fig. 2A are sketched for the predominant antiparallel sheet. When cutting the cylindrical barrel at the place where the first strand reaches the upper barrel end and flattening it out, the relationship between the number of strands n, the shear number S, and the tilt angle 0t becomes obvious:

R = [(Sa) 2 -1- (nb)2]°'5/2yr tanc~ = Sa/nb R = nb/2yr cos0t

Negative values of the shear number S are observed in the extreme cases of left-handed fl-helices (Table I) and gramicidin-A artifacts (Table II). In canonical fl-barrels S is positive and ranges between n and n + 4, allowing for an optimum fl-sheet twist. TolC with S = n + 8 is an exception. Furthermore, S is always an even number because after running around the barrel, ridges and valleys of the pleated sheet have to be joined to ridges and valleys again. In other words, the hydrogen bonds repeat only every second residue. The relationship between n, S, ~ and R in fl-barrels is shown graph- ically in Fig. 2B and in Fig. 3. The smallest barrel considered has six strands. Two strongly distorted copies of it were found in chymotrypsin (Table I). Eight-stranded fl-barrels are more regular and much more common; the most prominent example is the TIM barrel. Among the water-soluble proteins a series of eight-stranded barrels exists with shear numbers ranging from 8 to 12 in TIM, streptavidin, and lipocalins, respectively. In all these cases the barrel interior contains a hydropho- bic core. In TIM barrels the active centers are invariably found at the carboxy-terminal end of the barrel. Streptavidin binds biotin at one barrel end. The same applies for lipocalins where the bound large nonpolar compounds reach down to the barrel center. The increasing ligand sizes and binding site depths from TIM to the lipocalins correspond to the increasing barrel radii caused by larger shear numbers. In contrast to the smaller/5-barrels of water-soluble proteins, those of membrane-inserted fl-barrels start at eight strands and run up to 22 (Table II). Presumably, the required tightly packed nonpolar barrel core of the water-soluble proteins limits the radius of circular barrels to small values. In contrast, transmembrane barrels form polar cores the stability of which depends on hydrogen bonds rather than on geometrically exact nonpolar packing contacts. Such polar cores can be constructed much more leisurely and may also include water molecules that increase TRANSMEMBRANE fl-BARRELPROTEINS 5 7 the interior volume. Accordingly, the interiors of the barrels with up to 12 strands are polar and solid, except for TolC with its exceptionally large shear number of 20. The large shear number gives rise to a large barrel radius (Fig. 3) which causes TolC to form a channel. The same applies for 0t-hemolysin, which, however, has a smaller shear number but two more ~-strands. Channels are of course also formed by all porins. A general porin contains 16 fl-strands and has a shear number of 20 and a nearly circular cross section (Table II). Three parallel barrels associate to form trimers. The type of residues outlining the channel determines the specificity of a porin which, however, is usually not very strict. The two 18-stranded porins are very specific. Their channel cross-sections are actually smaller than those of the general porins in agreement with their higher selectivity. The 22-stranded barrels of the iron transporter proteins have circular cross sections and would form a very wide channel if they were not filled with the globular N-terminal 150-residue domain. Apart from the construction principles dictated by the fl-barrel geometry and illustrated in Fig. 2, the transmembrane fl-barrels follow additional rules that are probably dictated by factors other than the co- valent peptide structure:

1. Both the N- and C- termini are at the periplasmic barrel end re- stricting the strand number n to even values. 2. All/3-strands are antiparallel and locally connected to their next neighbors. 3. On trimerization a nonpolar core is formed at the molecular threefold axis of the porins so that the central part of the trimer resembles a water-soluble protein. 4. The external fi-strand connections are long loops named L1, L2, etc., whereas the periplasmic strand connections are generally minimum length turns named T1, T2, etc. 5. Cutting the barrel as shown in Fig. 2 and placing the periplasmic end at the bottom, the chain runs from the right to the left. 6. In all porins the constriction at the barrel center is formed by an inserted long loop L3. 7. The fl-barrel surface contacting the nonpolar membrane interior is coated with aliphatic side chains forming a nonpolar ribbon. The two rims of this ribbon are lined by girdles of aromatic side chains. 8. The sequence variability in transmembrane ~-barrels is higher than in water-soluble proteins and exceptionally high in the external loops. 58 GEORGE. SCHULZ

The first three rules are likely to reflect the folding process of the trimeric porins. Presumably, the central part folds in the periplasm like a water-soluble protein. The membrane-exposed parts of the barrels are then formed on insertion into the membrane. The short turns of rule 4 may facilitate barrel formation inside the membrane. Presumably rule 5 is a consequence of rule 4 because appropriate short turns can only be formed in one of the two possible chain directions. Rule 6 seems to reflect an early evolutionary event that has not yet been revised. The aromatic girdles of rule 7 were suggested to stabilize the t-barrel and its vertical position in the membrane (Schulz, 1992). The stabilization has been confirmed experimentally by demonstrating the preference of aromatic compounds for the two nonpolar-polar transition regions of the membrane (Wimley and White, 1996; Yau et al., 1998). Rule 8 came as a surprise to those with a high respect for membrane proteins which, of course, is mainly caused by our difficulties in solving membrane protein structures. Rule 8 explains these difficulties because it indicates that membrane proteins are subjected to fewer structural restraints than water-soluble ones and for this reason in general are more mobile and thus less crystallizable. Given the drastic mobility differences between the external loops and the membraneous and periplasmic moieties of the barrels, it was suggested that these proteins can be crystallized by creating suitable packing contacts through semirandom mutagenesis at loops and turns (Pautsch et al., 1999). Without structural knowledge, these loops and turns can often be predicted from the sequence and from interaction studies. For OmpA and OmpX this approach resulted in surprising successes. The procedure is also applicable to a-helical membrane proteins inasmuch as their crystallization problem is governed by their small polar surfaces. It is quite possible, however, that they hesitate to crystallize because their transmembrane a-helices are stabilized by the native laminar membrane environment and become flexible in the detergent/lipid micelles used for crystallization. With so many rules, the prediction of transmembrane t-barrels from the sequence should be achievable at a high confidence level. However, the simple approach of looking for alternating polar and nonpolar residues inside and outside the barrel is not very helpful because this pattern is frequently broken by nonpolar residues on the inside. Moreover, the t-strands are merely slightly more than half a dozen residues long which limits their information content appreciably. These problems have been tackled in several prediction programs (Welte et al., 1991; Schirmer and Cowan, 1993; Gromiha et al., 1997; Seshadri et al., 1998; Diederichs et al., 1998; Jacoboni et al., 2001) but cannot TRANSMEMBRANE fl-BARREL PROTEINS 59 be considered solved. For some time to come, the safety of t-barrel prediction will remain well below that of transmembrane 0t-helices with their nonpolar 25-residue segments.

IV. FUNCTIONS OmpX is synthesized in large amounts in stress situations and it is probably used as a defensive weapon binding to and thus interfering with foreign proteins. Half of the OmpX/%barrel protrudes into the external medium, presenting an inclined/%sheet edge that binds to any foreign protein with a/%strand in its surface layer (Vogt and Schulz, 1999). Such proteins are ubiquitous, an example being the large group of proteins with central parallel/%sheets surrounded by u-helices. In X-ray analysis, all loop residues of OmpX were located in the electron density distribu- tion indicating that the/%sheet edge presented to the foreign proteins is rigid, as it is required for tight binding. In contrast to OmpX the long external loops of OmpA are highly mobile and, on the whole, not visible in the respective electron density map. In vivo, the mobile loops are rather resistant to proteolytic attack, presumably because they bind to the surrounding lipopolysaccharides. Obviously, the mobile loops fulfill essential functions in bacterial life (Morona et aL, 1984). They are also known as docking points tbr bacteriophages. Among the outer membrane enzymes, OmpT is a special protease that has been implicated in the pathogenicity of bacteria. It is monomeric with the active center pointing to the outside (Vandeputte-Rutten et al., 2001). Another enzyme, the phospholipase A OmpLA, produces holes in the outer membrane when it is activated. The activation process has not yet been clarified, but it is known to require a dimerization of OmpLA in the membrane. The activation by dimer formation has been verified by a crystal structure analysis of an OmpLA dimer that was produced by a re- action with an inhibitor (Snijder et al., 1999). It showed that each active center contained a catalytic triad Ser-His-Asn on one subunit and an ox- anion hole formed by an amide together with a hydrated Ca 2+ ion on the other. The active centers are well placed for deacylating lipopolysaccharides of the external leaflet of the outer bacterial membrane. OmpLA functions in the secretion of colicins and virulence factors. All general porins contain pores with sizes allowing the permeation of molecules up to molecular masses of about 600 Da (Nikaido, 1994). The pores come with various selectivities. The porin from Rhodobacter capsulatus, for instance, contains a rather nonpolar binding site near the external end of the pore eyelet, indicating that it may pick up molecules 60 GEORG E. SCHULZ such as adenosine at very low concentrations. The structure also revealed a transversal electric field across the pore eyelet that acts as a polarity sep- arator, excluding the unwanted nonpolar compounds (Schulz, 1992). Porin OmpF from Escherichia coli has been thoroughly analyzed by numerous groups and became the first membrane protein to form X-ray grade crystals (Garavito and Rosenbusch, 1980). It is closely homologous to the porins PhoE (Cowan et al., 1992) and OmpC. These three porins show permeation properties adjusted to different environmental conditions. PhoE allows an efficient uptake of phosphate. OmpC from E. coli and its homologue OmpK36 from Klebsidla pneumoniae (Dutzler et al., 1999) are osmoporins expressed at high osmotic pressures usually caused by high salt concentrations. The high salt OmpC differs from the low salt OmpF mainly by an increased number of charged residues pointing into the pore lumen. Presumably, the charge increase counter- acts Debye-Hfickel shielding at high ionic strength so that OmpF and OmpC (OmpK36) have comparable permeation properties in differing environments. Additional structures were established for the main porin from Paracoccus denitrificans (Hirsch et aL, 1997) and for Omp32 from Comomonas acidovorans (Zeth et al., 2000). They confirmed the established general features of their homologues. All structurally established porins are aggregates of three parallel /5-barrels, each of which contains a single polypeptide chain. The /5- barrels contain either 16 or 18 strands. The interfaces are usually large and tightly packed. Therefore, it is not conceivable that the subunits are stable as monomers, either in the membrane or in the periplasm. How- ever, the existence of a functional monomeric porin has been reported (Conlan et aL, 2000), but its structure has not yet been elucidated. With maltoporin the bacteria developed a particular small pore that is adapted to the amylose helix and accepts only glucose units (Schirmer et al., 1995; Meyer et al., 1997). The ene rgetics of a maltooligosaccharide diffusing through such a pore has been examined in detail, revealing a combination of nonpolar and optimally spaced polar interactions that result in smooth gliding (Meyer and Schulz, 1997). This energy profile has been confirmed in a more detailed molecular dynamics study (Schirmer and Phale, 1999) and in an experimental study based on mutants (Dumas et al., 2000). The sucrose porin is a homologue of the maltoporin and very specific for the small molecule sucrose, in contrast with the maltoporin specific for larger oligomers (Forst et al., 1998). Besides porin trimers with 16- and 18-stranded/3-barrels, even larger 22-stranded /5-barrel proteins were found in the outer membrane, namely the monomeric active iron transporters FhuA and FepA. The lack of ATP or an equivalent energy carrier in the periplasm restricts the TRANSMEMBRANEfl-BARREL PROTEINS 61 outer membrane, in the first place, to more or less specific but passive pores that are not able to transport any solute against a concentration gradient. Bacteria overcame this problem by inventing plugged fi-barrels and the TonB apparatus. The structure of the two evolutionarily related plugged pores FhtLA and FepA are known. Their /%barrels have diameters of about 40 A (Table II). They bind siderophore-encapsulated iron in the external half of the barrel and are obstructed by an N-terminal 150-residue domain in the periplasmic barrel half. Mutational studies revealed their TonB binding sites. The directed iron transport through the outer membrane is energized by an interaction with TonB of the inner membrane that can draw energy from the cytosolic ATP pool (Rutz et al., 1992; Larsen et al., 1999). The plug formed by the 150-residue domain is removed after binding to TonB, making the siderophore available for internalization. A completely different principle has been followed by the heptameric 0t-hemolysines (Song et al., 1996; Olson et al., 1999). These proteins associate with their extramembrane domains. Subsequently, each subunit donates a/~-hairpin to form a common 14-stranded t-barrel through the membrane. In a similar manner TolC is assembled from three a-helical subunits (Koronakis et al., 2000). The subunits form a long, wide channel that spans the periplasm in their a-helical part and that is prolonged through the outer membrane by the t-barrel. The channel is used for the export of xenobiotics. According to the endosymbiotic theory, the outer membrane of gram- negative bacteria corresponds to the outer membranes of mitochondria and chloroplasts (Reumann et al., 1999). All of them are porous and cannot hold an electric potential difference. One long-term candidate for a porin homologue is the voltage-dependent anion channel (VDAC) of the outer mitochondrial membrane (Mannella, 1997). A further candidate for a t-barrel channel in the outer mitochondrial membrane is Tom40, which contains t-structure and forms a pore (Hill et al., 1998). Its molecular mass would suggest a t-barrel of the size of general porins. Unfortunately, none of these proteins has as yet yielded crystals suitable for structure analysis. Presumably, they are particularly difficult to crystallize because they face the soft cytosol, which does not require tough structures, in contrast to their bacterial counterparts that face the external medium demanding much higher stability.

g. FOLDING AND STABILITY On first sight the folding process of t-proteins appears to be much more complex than that of a-proteins. This does not apply, however, for 62 GEORG E. SCHULZ the all-next-neighbor antiparallel transmembrane t-barrels discussed here. For the porins it was suggested that the central part of the ho- motrimer including all N- and C-termini folds in the periplasm like a water-soluble protein so that the membrane-facing parts of the t-barrels dangle as 200-residue loops into the solvent (Schulz, 1992). On membrane insertion, these loops can then easily meander forming the special t-sheet topology. The simplicity of the folding process is corroborated by the fact that porins and other transmembrane t-barrels such as OmpA (Pautsch and Schulz, 1998) can be (re)natured from inclusion bodies. This production method worked even for the large monomeric/3-barrel of FepA. In vivo, the folding process may be supported by a periplasmic chaperone called Skp. Skp is a 17 kDa protein associated with the plasma membrane that, together with peptidyl prolyl isomerases and disulfide- exchanging enzymes, helps folding freshly synthesized proteins in the periplasm (Sch/ffer et al., 1999). Skp binds to partially unfolded polypep- tides. Depending on the presence ofphospholipids, lipopolysaccharides, and bivalent cations, Skp exists in two conformations, one of which is protease-sensitive (DeCock et al., 1999). Moreover, it was shown that Skp binds to unfolded periplasmic proteins and inserts into phospho- lipid monolayers, corroborating its putative role as helper in folding and membrane insertion. The stability of the t-barrel itself was demonstrated in engineering experiments with OmpA. The four external loops of OmpA were replaced by shortcuts in all possible combinations (Koebnik, 1999). The resulting deletion mutants lost their biological functions in bacterial F-conjugation and as bacteriophage receptors, but kept the transmembrane t-barrel as demonstrated by their resistance to proteolysis and thermal denaturation. The results confirm the expectation that the large external loops do not contribute to t-barrel folding and stability. Less dramatic changes were applied to a-hemolysin where the t-strand sequence was altered by reversing the sequence within the t-hairpin contributed by each subunit to the t-barrel (Cheley et aL, 1999). With respect to the t-barrel this changed only the hydrogen bonding pattern, but it reversed the sequence in the t-turn at the end of the hairpin which should, therefore, have local conformational consequences. It turned out that the "retro"-barrel formed a channel but failed to invade erythrocytes. A high activity could be obtained, however, when the t-turn was left in its original amino acid sequence, demonstrating that the tight t-turns at one end of these barrels are important for the stability of the whole barrel. Unfortunately, the detailed structures of the "retro"-barrels remained unknown. TRANSMEMBRANE fl-BARREL PROTEINS 63

The interiors of the three small t-barrels of OmpA, OmpX, and OmpLA are filled with polar residues forming a hydrogen bonding network. A number of separated cavities mostly filled with water have been reported for OmpA and OmpX. Accordingly, these/3-proteins are rigid inverse micelles and unlikely to form pores. For OmpA there was a lengthy discussion on the question of pore formation because small pores had actually been detected by ion fluxes (Sugawara and Nikaido, 1994; Arora et al., 2000). It seems likely that these data were obtained with OmpA preparations that had lost the internal barrel structure during the protein purification process permitting channel formation. Such conformations should differ appreciably from the crystalline structure. Using an atomic force microscope (AFM) operated in the tapping mode, the surface structure of OmpF was analyzed under several conditions. For this purpose, two-dimensional crystals of the porin OmpF from E. coli were mounted on graphite or mica and exposed to large voltages and low pH as well as high ionic strength environments (Mfiller and Engel, 1999). When operating in the contact mode at minimum force the AFM showed a large conformational change in the external loops reminiscent of a structural collapse. The authors relate this to voltage closure. Unfortunately, the experimental setup is somewhat artifical and the actual voltage across the porin cannot be measured. However, the effects are obvious and certainly report a structural weakness in the pro- trusion formed by the external loops. Since the AFM works with living material and is able to detect vertical differences in the range of atomic bond lengths, it is likely to yield most interesting results in the future. A pore feature attracting continuous interest is voltage gating. This effect remains to be explained in detail. It is observed at comparatively high voltages across the membrane. Since the corresponding electric field strength is so high that it may disrupt hydrogen bonds, it seems likely that most of the in vitro voltage gating studies (Liu and Delcour, 1998; Saxena et al., 1999) report nothing else than a structural breakdown inside a pore. This does not apply to the voltage-dependent anion channel of the outer mitochondrial membrane (VDAC), however, which shows channel closure in vivo (Mannella, 1998). Presumably, VDAC contains a solid but separate mobile and charged domain that can be driven by the electric field onto the pore so that it prevents any further ion flow.

WI. CHANNELENGINEERING The diffusion of small solutes through porins is passive. The diameters of the pore eyelets range from 10 • for the general porins to 6 for the highly selective porins. Larger pores are usually decorated with 64 GEORG E. SCHULZ oppositely charged residues at opposite sides that form a local transversal electric field at the pore eyelet. This field constitutes an energy barrier for low-polarity solutes (Schulz, 1992) so that the bacterium can exclude the unwanted nonpolar molecules such as antibiotics despite presenting a spacious eyelet for picking up large polar molecules such as sugars. The engineering of porins became popular after it had been demonstrated that a mass-produced porin (re)natured from inclusion bodies had assumed exactly the native conformation (Schmid et al., 1996). A systematic study changing the pore properties by mutations showed a strong correlation between eyelet cross section and diffusion rate (Schmid et al., 1998). Furthermore, a series of nine porins with mutations at the eyelet was analyzed with respect to ion conductance, ion selectivity, and voltage gating (Saxena et al., 1999). It was shown that charge reversals affect selectivity and voltage gating. Similar results were obtained with mutation at loop L3 of PhoE (VanGelder et al., 1997). In contrast to modifications at loop L3 inside the t-barrel, mutations at barrel wall residues lining the eyelet had only minor effects on voltage gating. This corroborates the suggestion that voltage gating reflects a structural breakdown in the pore. Sucrose porin has a somewhat larger pore than its homologue maltoporin, the pore eyelet of which is closely adjusted to a (1 ~ 4)-bound glucose units (Meyer and Schulz, 1997). With mutations at loop L3 the specificity of the sucrose porin was changed toward that ofa maltoporin (Ulmke et al., 1999). The specificity change was achieved by introduc- ing three eyelet-defining residues of the maltoporin and by removing the additional N-terminal 70-residue domain of the sucrose porin, the structure of which is not yet known. The width of a passive channel can be monitored by ionic currents through a black-lipid membrane harboring the respective membrane protein. The channel may be clogged by organic molecules reducing the current over the residence time of such a molecule. This time as well as the amount of the current reduction is characteristic for the applied compound. Gu et al. (1999) used this principle by placing a cy- clodextrin as an adapter into the 14-stranded fi-barrel of an engineered a-hemolysin. Using this device they determined the current reductions and the residence times caused by a number of modified adamantans, demonstrating that they can measure these compounds in concentrations of around 10/zM and distinguish between various types. In a further study Braha et al. (1997) introduced a zinc ion binding site into the t-barrel of a-hemolysin. When occupied, this site changed the channel conductivity, so that this barrel could be used as a zinc ion sensor. These TRANSMEMBRANE fl-BARREL PROTEINS 65 are marvelous engineering examples, converting a/%barrel pore 1o a molecular sensor with an electric output signal. In addition to the manifold possibilities of engineering on native /%barrels, these can also be designed from scratch. In an ab initio approach Ghadiri et al. (1994) designed cyclic octapeptides and demonstrated that these assemble to socalled "nanotubes" forming channels through a membrane. The octapeptides consisted of alternating D- and L-amino acid residues and thus followed closely the construction principle ofgramicidin A. In its native conformation gramicidin A forms a single strand/%barrel (n = 1) with a shear number S = 6 that may also be called a/3-helix. The nanotubes are close to the same construction with n = 1 and a shear number S = '8', but forming a ring instead of a/%helix. Stacking these rings across the membrane in a E-barrel with an inclination angle c~ = 90 ° (Fig. 2A) then produces the channel.

VII. CONCLUSIONS Although the major fraction of transmembrane proteins is a-helical, their/3-barrel counterparts have become popular because many of them could be analyzed in atomic detail and at high resolution. The transmembrane/3-barrel proteins occur only in the outer membranes of bacteria, mitochondria, and chloroplasts. They assume astonishingly regular conformations giving rise to numerous rules for their construction. It has been suggested that these regularities are required by the folding process. They are likely to permit the detection of transmembrane/?-barrels from the sequence at a reasonable confidence level in the future. Whereas water-soluble proteins contain regular/3-barrels with up to eight strands and nonpolar interiors, transmembrane/~-barre!s contain eight or more strands and have polar interiors. The smaller transmembrane/%barrels have solid cores partially filled with water. Accordingly, they can be considered inverse micelles. The larger barrels of this type have channels along their axis that allow for the permeation of various types of solutes. The largest known 22-stranded transmembrane /3-barrels are used for the active transport of rare commodities through the bacterial outer membrane. Their interior contains a globular protein domain which functions as a plug that can be removed for transport. The success rate for engineering transmembrane/%barrels appears to be superior to that for soluble proteins. On one hand, these barrels can be mass produced into inclusion bodies and (re)natured fllerefrom in vitro. On the other hand, the interiors of the fl-barrels housing the permeation channels can be mutated with little effect on the barrel 66 GEORG E. SCHULZ construction. The instabilities usually introduced with amino acid residue changes can be compensated for to a large degree by the surrounding t-barrel. This allows the application of a rather broad spectrum of mutations without destroying the protein construction as with water-soluble proteins in which mutations are frequently punished by deterioration. Taking advantage of this situation, the t-barrels have already been engineered to sensors for organic molecules and metal ions at micromolar concentrations, which, in the future, may lead to useful technical devices.

REFERENCES Arora, A., Rinehart, D., Szabo, G., and Tamm, L. K. (2000). Refolded outer membrane protein A ofEscherichia coliforms ion channels with two conductance states in planar lipid bilayers.J. Biol. Chem. 275, 1594--1600. Arora, A., Abildgaard, E, Bushweller, J. H., and Tamm, L. K. (2001). Structure of outer membrane protein A transmembrane domain by NMR spectroscopy. Nat. Struct. Biol. 8, 334-338. Banner, D. W., Bloomer, A. C., Petsko, G. A., Phillips, D. C., Pogson, C. I., Wilson, I. A., Corran, P. H., Furth, A. J., Milman, J. D., Offord, R. E., Priddle, J. D., and Waley, S. G. (1975). Structure of chicken muscle triose phosphate isomerase determined crystallographicallyat 2.5 h resolution using amino acid sequence data. Nature 255, 609-614. Baumann, U., Wu, S., Flaherty, K. M., and McKay, D. B. (1993). Three-dimensional structure of the alkaline protease of Pseudomonas aeruginosa: a two-domain protein with a calcium binding parallel beta roll motif. EMBOJ. 12, 3357-3364. Birktoft, J.J., mad Blow, D. M. (1972). Structure of crystalline ~-chymotrypsin.J. Mol. Biol. 68, 187-240. Bowie, J. U. (1999). Helix-bundle membrane protein fold templates. Protein Sci. 8, 2711- 2719. Braha, O., Walker, B., Cheley, S., Kasianowicz, J. J., Song, L., Gouaux, J. E., and Bayley, H. (1997). Designed protein pores as components for biosensors. Chem. Biol. 4, 497-505. Brenner, S. E., and Levitt, M. (2000). Expectations from structural genomics. Protein Sci. 9, 197-200. Buchanan, S. K., Smith, B. S., Venkatramani, L., Xia, D., Esser, L., Palnitkar, M., Chakraborty, R., van der Helm, D., and Deisenhofer, J. (1999). Crystal structure of the outer membrane active transporter FepA from Escherichia coli. Nat. Struct. Biol. 6; 56-63. Cheley, S., Braha, O., Lu, X., Conlan, S., and Bayley, H. (1999). A functional protein pore with a "retro" transmembrane domain. Protein Sci. 8, 1257-1267. Conlan, S., Zhang, Y, Cheley, S., and Bayley, H. (2000). Biochemical and biophysical characterization of OmpG: a monomeric porin. Biochemistry 39, 11845-11854. Cowan, S. W., Schirmer, T., Rummel, G., Steiert, M., Ghosh, 1L, Pauptit, R. A.,Jansonius, J. N., and Rosenbusch, J. P. (1992). Crystal structures explain functional properties of two E. coli porins. Nature 358, 727-733. DeCock, H., Schiller, U., Potgeter, M., Demel, R., Mfiller, M., and Tommassen,J. (1999). Affinity of the periplasmic chaperone Skp of Escherichia coli for phospholipids, lipopolysaccharides and non-native outer membrane proteins. Eur. J. Biochem. 259, 96-103. TRANSMEMBRANEfl-BARREL PROTEINS 67

Diederichs, K., Freigang, J., Umhau, S., Zeth, K., and Breed, J. (1998). Prediction by a neural network of outer membrane/3-strand protein topology. Protein Sci. 7, 2413- 2420. Dumas, E, Koebnik, R., Winterhalter, M., and Van Gelder, E (2000). Sugar transport through maltoporin of Escherichia coli.J. Biol. Chem. 275, 19747-19751. Dutzler, R., Rummel, G., Alberti, S., Hernfmdez-All6s, S., Phale, E S., Rosenbusch, J. E, Benedi, V.J., and Schirmer, T. (1999). Crystal structure and functional characterization of OmpK36, the osmoporin of Klebsiella pneumoniae. Structure 7, 425-434. Ferguson, A. D., Hofmann, E., Coulton, J. W., Diederichs, K., and Welte, W. (1998). Siderophore-mediated iron transport: crystal structure of FhuA with bound lipopolysaccharide. Science 282, 2215-2220. Fernfindez, C., Adeishvili, K., and Wfithrich, K. (2001a). Transverse relaxation-optimized NMR spectroscopy with the outer membrane protein OmpX in dihexanoyl phos- phatidylcholine micelles. Proc. Natl. Acad. Sci. USA 98, 2358-2363. Fern£mdez, C., Hilty, C., Bonjour, S., Adeishvili, K., Pervushin, K., and Wfithrich, K. (2001b). Solution NMR studies of the integral membrane proteins OmpX and OmpA from Escherichia coll. FEBS Lett. 504, 173-178. Forst, D., Welte, W., Wacker, T., and Diederichs, K. (1998). Structure of the sucrose- specific porin ScrY from Salmonella typhimurium and its complex with sucrose. Nat. Struct. Biol. 5, 37-46. Garavito, R. M., and Rosenbusch, J. P. (1980). Three-dimensional crystals of an integral membrane protein: an initial X-ray analysis. J. Cell Biol. 86, 327-329. Ghadiri, M. R., Granja, J. R., and Buehler, L. K. (1994). Artificial transmembrane ion channels from self-assembling peptide nanotnbes. Nature 369, 301-304. Gromiha, M. M., Majumdar, R., and Ponnuswamy, P. K. (1997). Identification of membrane spanning/3-strands in bacterial porins. Protein Eng. 10, 497-500. Gu, L. Q., Braha, O., Conlan, S., Cheley, S., and Bayley, H. (1999). Stochastic sensing of organic analytes by a pore-forming protein containing a molecular adapter. Nature 398, 686-690. Hendrickson, W. A., P~thler, A., Snfith, J. L., Satow, Y., Merritt, E. A., and Phizackerley, R. P. (1989). Crystal structure of core streptavidin determined from multi-wavelength anomalous diffraction of synchrotron radiation. Proc. Natl. Acad. SCI. USA 86, 2190- 2194. Hill, K., Model, K., Ryan, M. T., Dietmeier, K., Martin, E, Wagner, R., and- Pfanner, N. (1998). Tom40 forms the hydrophilic channel of the mitochondrial import pore for preproteins. Nature 395, 516-521. Hirsch, A., Breed, J., Saxena, K., Richter, O. M. H., Ludwig, B., Diederichs, K., and Welte, W. (1997). The structure of porin from Paracoccus denitrificans at 3.1 A resolution. FEBS Lett. 404, 208-210. Jacoboni, I., Martelli, E L., Fariselli, P., de Pinto, V., and Casadio, R. (2001). Prediction of the transmembrane regions of/Lbarrel membrane proteins with a neural network- based predictor. Protein Sci. 10, 779-787. Jap, B. IL (1989). Molecular design of PhoE porin and its functional consequences. J. Mol. Biol. 205, 407-419. Ketchem, R. R., Hu, W., and Cross, T. A. (1993). High-resolution conformation of gramicidin A in a lipid bilayer by solid-state NMR. Science 261, 1457-1460. Koebnik, R. (1999). Structural and functional roles of the surface-exposed loops of the /3-barrel membrane protein OmpA from Escherichia coli.J. Bacteriol. 181, 3688-3694. Koronakis, V., Sharff, A., Koronakis, E., Luisi, B., and Hughes, C. (2000). Crystal structure of the bacterial membrane protein TolC central to multidrug efflux and protein export. Nature 405, 914-919. 68 GEORG E. SCHULZ

Kreusch, A., and Schulz, G. E. (1994). Refined structure of the porin from Rhodopseu- domonas blastica: comparison with the p0rin from Rhodobacter capsulatus. J. Mol. Biol. 243, 891-905. Langs, D. A. (1988). Three-dimensional structure at 0.86 A of the uncomplexed form of the transmembrane ion channel peptide gramicidin A. Science 241, 188--191. Larsen, R. A., Thomas, M. G., and Postle, IC (1999). Protonmotive force, ExbB and ligand- bound FepA drive conformational changes in TonB. Mol. Microbiol. 31, 1809-1824. Liu, J., and Rost, B. (2001). Comparing function and structure between entire proteomes. Protein Sei. 10, 1970-1979. Liu, N., and Delcour, A. H. (1998). The spontaneous gating activity of OmpC porin is affected by mutations of a putative hydrogen bond network or of a salt bridge between the L3 loop and the barrel. Protein Eng. 11,797-802. Liu, W. M. (1998). Shear numbers of protein t-barrels: definition, refinements and statis- tics.J. Mol. Biol. 275, 541-545. Locher, K. P., Rees, B., Koebnik, R., Mitschler, A., Moulinier, L., Rosenbusch, J. E, and Moras, D. (1998). Transmembrane signaling across the ligand-gated FhuA receptor: crystal structures of free and ferrichrome-bound states reveal allosteric changes. Cell 95, 771-778. Mannella, C. A. (1997). On the structure and gating mechanism of the mitochondrial channel, VDAC.J. Bioenerg. Biomembr.. 29, 525-531. Mannella, C. A. (1998). Conformational changes in the mitochondrial channel protein, VDAC, and their functional implications.J. Struct. Biol. 121,207-218. Meyer, J. E. W., and Schulz, G. E. (1997). Energy profile of maltooligosaccharide permeation through maltoporin as derived from the structure and from a statistical analysis of saccharide-protein interactions. Protein Sci. 6, 1084-1091. Meyer, J. E. W., Hofnung, M., and Schulz, G. E. (1997). Structure of maltoporin from Salmonella typhimurium ligated with a nitrophenyl-maltotrioside.J.Mol. Biol. 266, 761- 775. Morona, R., Klose, M., and Henning, U. (1984). Escherichia coli K-12 outer membrane protein (OmpA) as a bacteriophage receptor: analysis of mutant genes expressing altered proteins.J. Bacteriol. 159, 570-578. Mfiller, D.J., and Engel, A. (1999). Voltage and pH-induced channel closure of porin OmpF visualized by atomic force microscopy. J. Mol. Biol. 285, 1347-1351. Murzin, A. G., Lesk, A. M., and Chothia, C. (1994a). Principles determining the structure of fl-sheet barrels in proteins. I. A theoretical analysis. J. Mol. Biol. 236, 1369-1381. Murzin, A. G., Lesk, A. M., and Chothia, C. (1994b). Principles determining the structure offl-sheet barrels in proteins. II. The observed structures.J. Mol. Biol. 236, 1382-1400. Nabedryk, E., Garavito, R. M., and Breton, J. (1988). The orientation of fl-sheets in porin. A polarized Fourier transform infrared spectroscopic investigation. Biophys. J. 53, 671-676. Newcomer, M. E.,Jones, T. A., Aqvist, J., Sundelin,J., Eriksson, U., Rask, L., and Peterson, P.A. (1984). The three-dimensional structure of retinol-bindingprotein. EMBOJ. 3, 1451-1454. Nikaido, H. (1994). Prevention of drug access to bacterial targets: permeability barriers and active effiux. Sc/ence264, 382-388. Olson, R., Nariya, H., Yokota, I~, Kamio, Y., and Gouaux, E. (1999). Crystal structure of staphylococcal LukF delineates conformational changes accompanying formation of a transmembrane channel. Nat. Struct. Biol. 6, 134-140. Pautsch, A., and Schulz, G. E. (1998). Structure of the outer membrane protein A transmembrane domain. Nat. Struct. Biol. 5, 1013-1017. TRANSMEMBRANEfl-BARREL PROTEINS 69

Pautsch, A., and Schulz, G. E. (2000). High resolution structure of the OmpA membrane domain. J. Mol. Biol. 298, 273-282. Pautsch, A., Vogt, J., Model, K., Siebold, C., and Schulz, G. E. (1999). Strategy for membrane protein crystallization exemplified with OmpA and OmpX. Proteins: Struct. Funct. Genet. 34, 167-172. Raetz, C. R. H., and Roderick, S. L. (1995). A left-handed parallel t-helix in the structurc of UDP-ALacetylglucosamine acyltransferase. Science 270, 997-1000. Reumann, S., Davila-Aponte, J., and Keegstra, K. (1999). The evolutionary origin of the protein-translocating channel of chloroplastic envelope membranes: identification of a cyanobacterial homolog. Proc. Natl. Acad. Sci. USA 96, 784-789. Rutz, J. M., Liu, J., Lyons, A.J., Goranson,J., Armstrong, S. K., McIntosh, M. A., Feix, J. B., and Klebba, E E. (1992). Formation of a gated channel by a ligand-specilic transport protein in the bacterial outer membrane. Science 258, 471-475. Saxena, K., Drosou, V., Maier, E., Benz, R., and Ludwig, B. (1999). Ion selectivity reversal and induction of voltage-gating by siteCirected mutations in the Paracoccus denitr!fi- cans porin. Biochemistry 38, 2206-2212. SchM'er, U., Beck, K., and Mfiller, M. (1999). Skp, a molecular chaperone of gram-negatiw~" bacteria, is required for the formation of soluble periplasmic intermediates of outer membrane proteins. J. Biol. Chem. 274, 24567-24574. Schirmer, T., and Cowan, S. W. (1993). Prediction of membrane-spanningt-strands and its application to maltoporin. Protein Sci. 2, 1361-1363. Schirmer, T., and Phale, E S. (1999). Brownian dynamics simulation of ion flow through porin channels.J. Mol. Biol. 294, 1159-1167. Schirmer, T., Keller, T. A., Wang, Y E, and Rosenbusch,J. E (1995). Structural basis tor sugar translocation through maltoporin channels at 3.1 ~, resolution. Science 267, 512-514. Schmid, B., Kr6mer, M., and Schulz, G. E. (1996). Expression of porin from Rhodopseu- domonas blastica in Escherichia coil inclusion bodies and folding into exact native strnc- ture. FEBS Lett. 381, 111-114. Schmid, B., Maveyraud, L., Kr6mer, M., and Schulz, G. E. (1998). Porin mutants with new channel properties. Protein Sci. 7, 1603-1611. Schulz, G. E. (1981). Protein differentiation: emergence of novel proteins during ew)lu- tion. Angew. Chem. Int. Ed. Engl. 20, 143-151. Schulz, G. E. (1992). Structure-function relationships in the membrane channel porin as based on a 1.8 ~ resolution crystal structure. In: Membrane Proteins: Structures, Interactions and Models (A. Pulhnan, J. Jortner, and B. Pullman, Eds.), pp. 403-412. Kluwer Academic, Dordrecht. Schulz, G. E. (1994). Structure-function relationships in porins as derived from a 1.8 ~\ resolution crystal structure. In: New Comprehensive Biochemistry: Bacterial Cell W~dl (J. M. Ghuysen, and R. Hakenbeck, Eds.), vol. 27, pp. 343-352. Elsevier, Amsterdam. Schulz, G. E., and Schirmer, R. H. (1979). P*inciples ofProtein Structure, pp. 84-87. Springel- Verlag, New York. Seshadri, K., Garemyr, R., Wallin, E., von Heijne, G., and Elofsson, A. (1998). Archi- tecture of t-barrel membrane proteins: analysis of trimeric porins. Protein Sci. 7, 2O26-2032. Sipos, L., and von Heijne, G. (1993). Predicting the topology of eukaryotic membrane proteins. Eur.J. Biochem. 213, 1333-1340. Snijder, H.J., Ubarretxena-Belandia, I., Blaauw, M., Kalk, K. H., Verhij, H. M., Egmond, M. R., Dekker, N., and Dijkstra, B. W. (1999). Structural evidence for dimerization- regulated activation of an integral membrane phospholipase. Nature 401, 717-721. 70 GEORGE. SCHULZ

Song, L., Hobaugh, M. R., Shustak, C., Cheley, S., Bayley, H., and Gouaux, J. E. (1996). Structure of staphylococcal a-hemolysin, a heptameric transmembrane pore. Science 274, 1859-1865. Sugawara, E., and Nikaido, H. (1994). OmpA protein ofEschetichia coli outer membrane occurs in open and closed channel forms.J. Biol. Chem. 269, 17981-17987. Ulmke, D., Kreth, J., Lengeler, J. W., Welte, W., and Schmid, T. (1999). Site-directed mutagenesis of loop L3 of sucrose porin ScrYleads to changes in substrate selectivity. J. Bacteriol. 181, 1920-1923. Vandeputte-Rutten, L., Kramer, R. A., Kroon,J., Dekker, N., Egmond, M. R., and Gros, E (2001). Crystal structure of the outer membrane protease OmpT from Escherichia coli suggests a novel catalytic site. EMBOJ. 20, 5033-5039. VanGelder, E, Saint, N., van Boxtel, R., Rosenbusch, J. E, and Tommassen, J. (1997). Pore functioning of outer membrane protein PhoE ofEscherichia coli: mutagenesis of the constriction loop L3. Protein Eng. 10, 699-706. Vogt, J., and Schulz, G. E. (1999). The structure of the outer membrane protein OmpX from Escherichia coli reveals possible mechanisms of virulence. Structure 7, 1301-1309. Wallace, B. A., and Ravikumar, K. (1988). The gramicidin pore: crystal structure of a cesium complex. Sc/ence241, 182-187. Weiss, M. S., and Schulz, G. E. (1992). Structure of porin refined at 1.8 A resolution. J. Mol. Biol. 227, 493-509. Weiss, M. S., Wacker, T., Weckesser, J., Welte, W., and Schulz, G. E. (1990). The three- dimensional structure of porin from Rhodobacter capsulatus at 3 A resolution. FEBS Left. 267, 268-272. Welte, W., Weiss, M. S., Nestel, U., Weckesser, J., Schiltz, E., and Schulz, G. E. (1991)." Prediction of the general structure of OmpF and PhoE from the sequence and structure of porin from Rhodobacter capsulatus. Orientation of porin in the membrane. Biochim. Biophys. Acta 1080, 271-274. Wimley, W. C., and White, S. H. (1996). Experimentallydetermined hydrophobicity scale for proteins at membrane interfaces. Nat. Stmct. Biol. $, 842-848. Yau, W. M., Wimley, W. C., Gawrisch, K., and White, S. H. (1998). The preference of tryptophan for membrane interfaces. Biochemistry 37, 14713-14718. Yoder, M. D., Keen, N. T., and Jurnak, E (1993). New domain motif: the structure of pectate lyase C, a secreted plant virulence factor. Sc/ence260, 1503-1507. Zeth, K., Diederichs, K., Welte, W., and Engelhardt, H. (2000). Crystal structure of Omp32, the anion-selectiveporin from Comamonas acidovorans, in complex with a periplasmic peptide at 2.1 A resolution. Structure 8, 981-992.